[SPARK-40872] Fallback to original shuffle block when a push-merged shuffle chunk is zero-size

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.3.0, 3.2.2
• Fix Version/s: 3.3.2, 3.4.0
• Component/s: Shuffle
• Assignee: gaoyajun02
• Reporter: gaoyajun02
• Created: 21/Oct/22 12:08
• Updated: 27/Nov/22 01:58
• Resolved: 21/Nov/22 06:14

Description

A large number of shuffle tests in our cluster show that bad nodes with chunk corruption appear have a probability of fetching zero-size shuffleChunks. In this case, we can fall back to original shuffle blocks.

Issue Links

• links to: [Github] Pull Request #38333 (gaoyajun02)
• links to: [Github] Pull Request #38751 (gaoyajun02)